Application of an income-based structural model to measure the probabilities of default of five european banks

  • Rafa Ben Yakhlef (Student)

Student thesis: Master's Thesis

Abstract

This dissertation analyses whether a modified version of the EBIT-based structural model by (Goldstein, Ju, & Leland, 2001) is able to replicate the default metrics published by major credit rating agencies in the case of banks. This research studies five European banks from 2001 until 2020. As the reference model focus on non-financial institutions, it was adapted to fit the characteristics of banks. In particular, the assumption that firms have fixed financial costs was replaced by the hypothesis that a fraction of banks’ non-interest costs are fixed. This share was determined in order to match credit rating agencies average probabilities of default, which equals 1.14% during our 20 years sample. After gathering all data, the model was calibrated following the iterative approach, first proposed by (Vassalou & Xing, 2004). A regression of the mean model probabilities of default and distances to default at each moment in time on the equivalent ratings-implied measures showed an R-squared of 0.27 and 0.40, respectively. Furthermore, this dissertation presents a panel data regression that assesses the fixed effects of each bank. The significance test shows that the coefficients in all regressions are significant at 5% significance levels except the fixed effects associated with three banks. I concluded that the model’s credit risk indicators are very comparable to the ratings given by credit rating agencies, though the correlation is far from perfect.
Date of Award27 Apr 2022
Original languageEnglish
Awarding Institution
  • Universidade Católica Portuguesa
SupervisorNuno Silva (Supervisor)

Keywords

  • Structural model
  • Banks
  • Credit ratings
  • Default prediction
  • Credit risk

Designation

  • Mestrado em Finanças

Cite this

'